There are any reason to use 'latin1' for encoding and decoding?

[dbwodlf3]

https://github.com/numba/llvmlite/blob/master/llvmlite/binding/common.py

In using llvmlite.binding.parse_assembly, There are some korean comments in ll file. So can't encode and decode.

If there aren't special reason to use 'latin1', i think it is better to use 'utf-8'.

Thanks.

[gmarkall]

UTF-8 was originally used, but was switched to latin1 in https://github.com/numba/llvmlite/pull/53. If I switch the encoding to UTF-8, the test case that accompanied that PR doesn't fail (on Linux at least, will try shortly on Windows). I'm not sure the reason for the switch to latin1 is still valid, so will post more back here after further investigation.

edit: corrected "latin8" to "latin1"

[gmarkall]

Update: The test also passes on Windows.

[gmarkall]

I've opened https://github.com/numba/llvmlite/pull/655 with the change to see what happens on CI.

[gmarkall]

The CI tests seem to pass with utf-8, so I've pinged @sklam to ask for a further opinion.

[gmarkall]

The linked PR #655 is now ready for review.